Research On Graph Privacy Protection Method Based On The Structural Entropy

The analysis and value extraction of large-scale network data can support scientific research such as decision-making,network security situational awareness,and user behavior prediction.Moreover,the measurement of network structural entropy(structural entropy)that supports the structural information embedded in real and complex physical systems,structural entropy can support the measurement of privacy information in networks and solve the most critical problem in privacy protection,which is significant to the study of graph theory.Accordingly,this study focuses on structural information theory.For instance,the study constructs a graph privacy protection model based on structural entropy in response to measuring structural information embedded in graphs and graph privacy protection needs.Furthermore,this study investigates the privacy measurement and privacy protection of graphs with complex topological structures based on the requirements of practical application scenarios,combined with the tree path similarity measurement theory,graph segmentation theory and node importance theory as research tools.Additionally,this study includes the similarity measurement of partition trees of graphs,the application of structural entropy theory in graph data privacy protection,the application of graph segmentation algorithm based on structural entropy in social network privacy control,the network security measurement model of P2 P social networks and node importance ranking based on network structural information.Therefore,the specific work of the study is as follows:(1)The research method of structural similarity of partition tree.Firstly,this study used the research method of structural similarity of partition tree to improve the similarity calculation method based on tree path to make it suitable for the partition tree of the highdimensional graph structure.Similarly,this study also proposed the relevant theorems and proved the theoretical feasibility.Secondly,according to the improved similarity calculation theory,a privacy measurement model of high-dimensional graph structure based on structure entropy is proposed.Finally,a data availability measurement model based on normalized mutual information is proposed for the privacy protection method of a two-dimensional graph structure.The model uses the characteristics of standardized mutual information to solve the problems of the change of graph structure data privacy information and the difficulty of data availability measurement after the change of community structure.(2)The privacy control scheme of improved graph segmentation algorithm based on structural entropy.The contract algorithm in the existing graph segmentation algorithm has some shortcomings in the clustering process,such as too random graph segmentation results and too significant computational overhead of subsequent privacy verification.Therefore,this paper proposes an improved contract algorithm privacy control scheme based on structural entropy,including two strategies: maximum probability information flow priority strategy(MIPS)and acquaintance priority strategy(APS)to reduce the uncertainty of segmentation results;In addition,referring to the concept of graph resistance,an algorithm utility evaluation mechanism based on graph resistance is established to evaluate the change of privacy protection algorithm to the ability of the network to resist privacy disclosure.(3)P2P social network security measurement model based on graph resistance.This study proposed a network security measurement model based on graph resistance to mitigate the current challenges,such as the lack of a specific P2 P network security measurement method and the difficulty of measuring the dynamic complexity of the existing distributed social network.Firstly,the dynamic complexity measurement model of P2 P network is designed in combination with structural entropy to solve the dynamic measurement problem caused by the high degree of freedom and high fault tolerance of P2 P social network.Secondly,the local feasibility and universality challenge the current network security measurement model.Therefore,this study constructed a network security measurement model based on a graph.Accordingly,the experimental and theoretical analyses show that the model supports global and local detection of network damage scope,and it is not limited by network size and type with a low computational cost.(4)Node importance index based on the structural information.The research on the importance of nodes is relatively mature,but most of schemes are only evaluated from a single perspective,and there are few schemes that comprehensively reflect the importance of nodes.Structural entropy can systematically represent the real state of the network and reflect the changes of the network caused by node changes.Therefore,through the combination of structure entropy method and node deletion method,the nodes are classified and ranked,and the node importance measurement index based on structural information is proposed.The experimental analysis shows that the node importance ranking method of fusion structure information is more accurate and comprehensive.